Add Padua points

[alexfikl]

This mostly just adds the Padua nodes for arbitrary orders (see wiki). Padua nodes are meant to be a unisolvent set of nodes for PN on the square with optimal growth of the Lebesgue constant.

Some questions / remarks:

• What would be some cool things to check?
• These points are not very symmetric. In particular, they have different numbers of nodes on faces. Will the face mass matrices still work?
• For the standard case, there's also an associated Chebychev-Gauss-type quadrature. Worth adding?
• The order of the nodes is a bit haphazard at the moment. Would be useful to sort them in lexicographic order like the tensor product case.

[inducer]

These points are not very symmetric. In particular, they have different numbers of nodes on faces. Will the face mass matrices still work?

Hmm. Probably. There's a face homogeneity assumption there currently which we would need to break. Fortunately, the data structures can take this. It's just the logic that needs some help.